摘要:针对以往SR电机多变量、多约束条件、多目标的设计寻优过程中容易陷入局部最优,并且电磁场有限元分析计算耗时长的缺点,本文提出一种能快速收敛实现多目标全局优化的电机设计方法。利用广义回归神经网络对目标函数进行非线性建模,通过模糊自适应粒子群算法找寻帕雷托最优解,实现了SR电机定转子磁极极弧的全局优化设计。研究结果表明:在电机主要尺寸不变的条件下,显著地降低了样机的转矩脉动,提高了平均输出转矩。
关键词:模糊自适应粒子群;平均转矩;转矩脉动;极弧;有限元;广义回归神经网络 Abstract: For the previous SR motor multi-variations, multi-constraints, multi-objective optimization design process is easy to fall into local optimum, and the shortcomings of electromagnetic finite element analysis calculation taking a long time, this paper proposed a multi-objective global optimization method with fast convergence Then the nonlinear objective functions were modeled by the generalized regression neural network, and fuzzy adaptive particle swarm optimization was used to find the Pareto optimal solution, finally the stator and rotor magnetic pole arc of SR motor achieved global optimization Results showed that under the main dimensions of the motor with the same conditions, the torque ripple of the prototype was significantly reduced and the average output torque was increased
Key words: fuzzy adaptive particle swarm; average torque; torque ripple; pole arc; finite element method; generalized regression neural network |